A few years ago when ML could only dream creepy dog faces, I would not believe that it could generate coherent images with decent framing and composition, corre

A few years ago when ML could only dream creepy dog faces, I would not believe that it could generate coherent images with decent framing and composition, correct proportions, realistic light and shadows, depth of field, excellent taste for palettes, and even good faces without uncanny valley — and hands and text being the hard parts. So this time I'm willing to believe that the small morphing clips and blurry NeRFs are going to grow into whole realistic 3d scenes.